Available for collaboration

Research Fellow · University of Siena

Leonardo
Guiducci

Applied AI researcher bridging Deep Reinforcement Learning and Computer Vision with real-world systems — from energy grids to clinical decision support.

Deep Reinforcement Learning Computer Vision Energy Management MLOps Healthcare AI Robotics
View projects ↓ LinkedIn ↗ GitHub ↗
01

About

I'm a Research Fellow at the Interaction Design Lab, University of Siena, working under Prof. Antonio Rizzo on AI applications for energy management and healthcare. My background is in Robotics Engineering (University of Perugia), which gives me a strong foundation in autonomous systems and algorithm development.

My research spans two main threads. In energy systems, I develop Deep Reinforcement Learning and federated learning approaches for microgrid control, Battery Energy Storage Systems, and renewable energy communities. In healthcare, I work on Computer Vision pipelines for clinical environments and dialogical AI systems for decision support.

I'm particularly interested in the gap between research prototypes and deployable systems — building things that actually run in hospitals, factories, and energy grids.

Selected for Google's CrescerAI initiative, recognising contributions to AI research in Italy.

Current position
Research Fellow · Univ. di Siena
Current project
FRESIA — Federated RL for renewable energy communities (FSE+)
Education
M.Sc. Robotics Engineering · Univ. di Perugia
Collaborations
Washington University in St. Louis · Karlstad University
02

Projects

☀️
Solar Forecasting MLOps
End-to-end MLOps pipeline for 24-hour solar power generation forecasting. Production-ready infrastructure with XGBoost, MLflow experiment tracking, Prefect orchestration, Evidently drift monitoring, and Grafana dashboards — all containerised with Docker Compose.

Capstone project for the MLOps Zoomcamp (DataTalksClub).
XGBoost MLflow Prefect Docker Grafana Evidently
🤖
E-VAT
Asymmetric End-to-End Deep Reinforcement Learning system for visual active exploration and target tracking in unknown environments. Uses an asymmetric actor-critic architecture with a Target-Detection Network and Exploration-Tracking Network.

Published in IEEE Robotics and Automation Letters, vol. 7, April 2022.
Deep RL Computer Vision PyTorch Robotics IEEE RA-L
SmartCer
Simulator for renewable energy communities developed as part of the SUMMA project, funded by Repubblica Digitale and a Google grant. Supports modelling of energy flows, storage scheduling, and community-level optimisation for smart grid applications.

See also: summa.unisi.it ↗
Energy Communities Simulation Python Smart Grid
03

Publications

2026
Applied Sciences
Dialogical AI for Cognitive Bias Mitigation in Medical Diagnosis
Guiducci, Saulle, Dimitri, Valli, Alpini, Tenti, Rizzo
2025
ESANN 2025
Introducing Intrinsic Motivation in Elastic Decision Transformers
Guiducci, Dimitri, Palma, Rizzo
2024
MECO 2024
Optimizing Energy Efficiency in Smart Factories: Battery Energy Storage Systems Integration Analysis
Guiducci, Palma, Rizzo
2023
MECO 2023
A Reinforcement Learning Approach to the Management of Renewable Energy Communities
Guiducci, Palma, Stentati, Rizzo, Paoletti
2022
IEEE RA-L
E-VAT: An Asymmetric End-to-End Approach to Visual Active Exploration and Tracking
Dionigi, Devo, Guiducci, Costante